Real-time route optimisation.
Vehicle routing is an old optimisation problem. Solving it continuously, on live event streams, is a much newer one.
What it is
With a couple of colleagues I built a real-time vehicle routing (VRP) optimisation platform on Microsoft Fabric's Real-Time Intelligence hub. Events flow in through Kafka and Azure Event Hubs, land in an Eventstream, get stored and queried in an Eventhouse with KQL, and surface on real-time dashboards with activators triggering downstream actions.
The project was a tour of essentially the entire Fabric Real-Time Intelligence stack. The team was selected to present the solution at Microsoft FabCon 2026 in Atlanta; I helped build the platform and advised on the presentation.
The work
- Co-built the streaming backbone: Kafka and Azure Event Hubs feeding Fabric Eventstream
- Eventhouse storage with KQL queries powering live views of the routing state
- Real-time dashboards and activators turning stream conditions into actions
- Continuous VRP re-optimisation on top of the live event flow
- Advised on the Microsoft FabCon 2026 Atlanta presentation of the solution
Concepts
What it taught me
Batch thinking dies hard. Designing for streams means accepting that the answer is always provisional, and building the system to embrace that.